yasin alan, jeffrey p. dotson, & m¨umin kurtulus¸ on the ... · category by developing new...

Yasin Alan, Jeffrey P. Dotson, & Mumin Kurtulus

On the Competitive and CollaborativeImplications of Category CaptainshipCategory captainship (CC) is a retailing practice wherein a retailer collaborates with one of the manufacturers in aproduct category (referred to as the captain) to develop and implement a categorymanagement strategy. AlthoughCChas been studied using both theoretical models and surveys, empirical evidence on the benefits and drawbacks of CCis scarce. The authors use a unique data set collected during aCC implementation to empirically examine the impact ofCC on the retailer, the captain, and the other manufacturers in the category. The authors find that both the retailer’sprivate label and the captain benefit from CC because of pricing and assortment changes. They also find that somecompeting manufacturers benefit from CC while others suffer. Specifically, the manufacturers that closely competewith the captain benefit, whereas the manufacturers that are in close competition with the private label suffer becausethe retailer protects its private label. The authors show that category sales would have been higher if the retailer hadnot protected its private label. This study sheds light on how joint consideration of assortment and pricing, the presenceof a private label, and product characteristics may influence the outcomes of CC implementations.

Keywords: category management, category captainship, private label, retailing, channel partnerships

Online Supplement : http://dx.doi.org/10.1509/jm.15.0196

Category management is a commonly used retailingpractice in which a retailer treats a product category(i.e., a set of similar products) as a strategic business

unit. A product category (e.g., canned vegetables, salty snacks,carbonated beverages) consists of products offered by nationalbrands and may also include private label products offered bythe retailer. Categorymanagement enables retailers to focus onmaximizing category performance, typically measured by thesales or profitability of the entire category, instead of makingdecisions on a product-by-product basis (Zenor 1994). Priorresearch inmarketing has shown that categorymanagement canbe beneficial for retailers because it enables them to simplify,coordinate, and thereby improve the process of makingassortment, pricing, and other merchandising decisions(ACNielsen 2005; Basuroy, Mantrala, and Walters 2001; Dhar,Hoch, and Kumar 2001).

Effective category management requires a retailer to alignits product offerings with evolving consumer needs. Becauseretailers manage many categories, constant monitoring andinterpretation of consumer trends is a costly and labor-intensive task for them. Manufacturers typically have a bet-ter understanding of consumer needs because their expertise

is focused on a much smaller set of products and categories(Blattberg and Fox 1995). The combination of retailers’ lack ofresources and manufacturers’ superior category knowledgecreates supply chain collaboration opportunities. Accordingly,many retailers manage some of their categories in collabo-ration with one of their leading manufacturers. These leadingmanufacturers are often referred to as category captains, andthe practice itself is referred to as category captainship (CC;Desrochers, Gundlach, and Foer 2003; Federal Trade Com-mission 2001, 2003).

Category captainship has become a preferred way ofexecuting category management. General Mills, for example,assisted one retailer in the dry packaged dinners category byreplacing slow-moving stockkeeping units with faster-turningproducts (Progressive Grocer 2011). Abbott Nutrition helpeda retailer in the baby food and consumables category by re-commending a new planogram with some new products in theassortment and changing prices to reflect the new productmix (ProgressiveGrocer 2010). In addition, J.M. SmuckerCo.helped several retailers in the canned and packaged beveragescategory by developing new shelf concepts and endcap dis-plays based on consumer insights (Progressive Grocer 2015).In summary, the captain’s recommendations vary acrossretailers and categories and may affect assortment, pricing,and/or merchandising decisions.

The trade literature has suggested that both retailers andmanufacturers can benefit from CC (e.g., Progressive Grocer2010, 2011, 2015). However, controversies regarding CC havearisen because the captain provides recommendations to theretailer regarding not only its own products but also those of itscompetitors. Consequently, the captainmay have a positive biastoward its own products to the detriment of the competitors’products. The term “competitive exclusion” has often been used

Yasin Alan is Assistant Professor of Operations Management, OwenGraduate School of Management, Vanderbilt University (e-mail: [email protected]). Jeffrey P. Dotson is Associate Professor ofMarketing and Global Supply Chain, Marriott School of Management,Brigham Young University (e-mail: [email protected]). Mumin Kurtulus‚ isAssociate Professor of Operations Management, Owen Graduate School ofManagement, Vanderbilt University (e-mail: [email protected]). The authors thank the JM review team for valuable feedbackthat greatly helped improve the quality of this article. Kusum Ailawadi servedas area editor for this article.

© 2017, American Marketing Association Journal of MarketingISSN: 0022-2429 (print) Vol. 81 (July 2017), 127–143

1547-7185 (electronic) DOI: 10.1509/jm.15.0196127

http://dx.doi.org/10.1509/jm.15.0196

mailto:[email protected]





http://dx.doi.org/10.1509/jm.15.0196

to refer to situations in which the captain uses its position to putits competitors at a disadvantage (Carameli 2004; Federal TradeCommission 2003).

The existing research on CC is based on legal theory (e.g.,Wright 2009); surveys (Gooner, Morgan, and Perreault 2011;Morgan, Kaleka, and Gooner 2007); game theoretic models ofretailer–manufacturer interactions under CC (e.g., Kurtulus andNakkas 2011; Kurtulus, Nakkas, and Ulku 2014; Subramanianet al. 2010); and structural estimation, which enables counter-factual analyses regarding how a hypothetical CC imple-mentation would have affected category decisions andperformance (Nijs, Misra, and Hansen 2014). However,empirical evidence on the collaborative and competitiveimplications of CC is scarce, as retailers are reluctant to shareCC data because of antitrust concerns (Nijs, Misra, andHansen 2014). From a collaborative standpoint, the existingliterature on CC does not provide a formal analysis of anactual CC implementation to assess whether and how CCbenefits the retailer and the captain. From a competitivestandpoint, there is no empirical evidence regarding whetherCC benefits or hurts the competingmanufacturers. Accordingly,our goal in this article is to empirically study the implicationsof CC for the retailer, captain, and competing manufacturersusing a unique data set collected during a CC implementation.

Our data set contains 52 weeks of product-level measuresfor all products in a shelf-stable food category with significantprivate label presence. During this time period, the retailerconducted a full category review in collaboration with one ofthe largest manufacturers in the category. Because the retailertreated this category as a revenue generator, the category reviewmainly focused on potential ways to increase category salesrevenue. Recommendations generated during this review ledto a new assortment and pricing strategy, which was imple-mented in week 21. By comparing sales revenue during the pre-and post-CC implementation periods (i.e., thefirst 20weeks andthe last 32 weeks in the data), we investigate the followingresearch questions:

1. Does the retailer benefit from CC? If it does, are the benefitsdriven by pricing or assortment changes? Are there any otherdrivers beyond assortment and pricing?

2.What is the impact of CC on different manufacturers in thecategory, including the private label and captain? Is it possiblefor the competing manufacturers (i.e., all manufacturersexcept the private label and captain) to benefit from CC?

3.What determines whether a competing manufacturer benefitsor suffers from CC?

Previous literature has raised similar questions (e.g.,Ailawadi et al. 2010), but data-driven answers are notreadily available. Addressing these questions through actualCC implementation data has several advantages over priorresearch on CC. First, the literature on CC considers pricingand assortment in isolation. For instance, Kurtulus and Nakkas(2011) andKurtulus et al. (2014) focus on howCC influences aretailer’s assortment, whereas Kurtulus and Toktay (2011) andNijs, Misra, and Hansen (2014) study how CC affects pricesfor a given assortment. While the existing models cannot fullycapture the CCphenomenon because of their focus on only onepossible lever, our study is unique in jointly considering pricingand assortment.

Second, many retailers view the private label as a keycomponent of a successful category strategy (Kumar andSteenkamp 2007). Accordingly, retailers often use private labelperformance as one of the metrics to evaluate category per-formance (ACNielsen 2005, Chapter 6). Despite its practicalrelevance and importance, the impact of CC on private labelproducts has not been considered in the existing CC literature.Our study shows howprivate label presencemay affect categorydecisions and performance in the CC context.

Third, the assortment literature in marketing has shown thatproducts with similar attributes (e.g., same size) are more likelyto compete for demand (e.g., Rooderkerk, Van Heerde, andBijmolt 2011, 2013). Following this literature, we specify anattribute-based demand model, whereas the existing CC liter-ature has used more stylized models (e.g., linear demandmodel). Our study sheds light on how a product’s similarity tothe captain’s products and private label products may affectwhether such a product benefits or suffers from CC.

In summary, while our empirical findings are based on oneCC implementation, our contribution stems from examiningseveral factors that have not been considered in theCC literature.In particular, our study informs practitioners and researchersby demonstrating how joint consideration of assortment andpricing, the presence of a private label, and product charac-teristics may influence the outcomes of CC implementations.

The rest of this article is organized as follows. First, wesummarize the relevant research and build a conceptual frame-work. We then describe our data set and empirical model. Next,we present our findings regarding the impact of CC. Finally, weconclude with a summary of our results and their implications aswell as a discussion of the limitations of our study.

Theoretical Background andConceptual Framework

In this section, we first provide a theoretical background for ourstudy by reviewing previous research on pricing, assortment,and merchandising in the context of category management.Because these literature streams are vast, we limit our attentionto studies that are most relevant to our setting. We then use thistheoretical background to develop our conceptual framework.

Theoretical Background

Retail pricing under category management. Shifting frombrand-centric management of retail prices to jointly settingprices in an entire category can lead to a significant improve-ment in category performance (e.g., Basuroy, Mantrala, andWalters 2001; Zenor 1994). Nonetheless, several factors makepricing a challenging task for retailers in the CC context. First,retailers should pay attention to cross-price effects because salesof a product depend on not only its own price but also the pricesof substitute products in the category (e.g., Besanko, Dube,and Gupta 2005; Kadiyali, Chintagunta, and Vilcassim 2000).Second, pricing decisions are influenced by the strategic role ofthe category (e.g., sales or profit maximization) for the retailer(ACNielsen 2005, p. 115). For instance, Zenor (1994) andBasuroy,Mantrala, andWalters (2001) show that implementingcategory management to maximize category profits can lead to

128 / Journal of Marketing, July 2017

higher retail prices compared with a brand-centric managementof retail prices. In contrast, Dhar, Hoch, andKumar (2001) showthat best-performing retailers in terms of category sales are theones with the lowest retail prices.

Third, the presence of a private label program is a key driverof a retailer’s pricing decisions (Chintagunta, Bonfrer, and Song2002).Many retailers give preferential treatment to private labelproducts (Kumar and Steenkamp 2007) because they typicallyhave higher percentage margins (Ailawadi and Harlam 2004),increase a retailer’s bargaining power relative to national brands(Chintagunta, Bonfrer, and Song 2002; Pauwels and Srinivasan2004), and can improve store loyalty in the case of high-qualityprivate label products (Corstjens and Lal 2000). Such a prefer-ential treatment affects retail prices. For instance, Chintagunta(2002) shows that the retailer’s desire to increase private labelmarket share drives the retail prices of private label productsbelow the levels obtained under a category profit-maximizationobjective. In addition, private label introduction may decreasethe retail prices of the national brands in the category (e.g.,Chintagunta, Bonfrer, and Song 2002; Du, Lee, and Staelin 2005).

Despite increasing a retailer’s bargaining power, increasedprivate label presence may not necessarily increase store trafficor revenues (Pauwels and Srinivasan 2004) and may lead tolower dollar margins per unit (Ailawadi and Harlam 2004).Moreover, pushing private labels too far may hurt store loyaltyas a result of an inverted U-shaped relationship between ahousehold’s private label share and store loyalty (Ailawadi,Pauwels, and Steenkamp 2008). In summary, private labelpresence exacerbates the difficulty of a retailer’s pricing deci-sions because it requires the retailer to retain a balance betweenprivate label and national brands.

Finally, transitioning from traditional category manage-ment, in which the retailer makes decisions on its own, to CCcan influence prices. The research studying this transition hasmodeled CC as an alliance between the retailer and the captainin settings inwhich the retailer tries tomaximize category profits(Kurtulus and Toktay 2011; Nijs, Misra, and Hansen 2014).These studies predict steep price decreases for the captain’sproducts because the formation of an alliance between theretailer and the captain mitigates double marginalization, whichenables the retailer to offer the captain’s products at significantlylower prices. The competing manufacturers, in contrast, con-tinue to suffer from double marginalization and make relativelyminor price reductions as a response to the downward pricepressure exerted by the captain (Kurtulus andToktay 2011;Nijs,Misra, and Hansen 2014). That is, the captainship literature hassuggested that CC leads to a decline in retail prices, with thesteepest price declines for the captain’s products.

Assortment planning under categorymanagement. Findingthe right assortment is a challenging task because most cate-gories have tens or even hundreds of products, which makes itdifficult for retailers to determine the right assortment size andcomposition. There is mixed evidence regarding the relation-ship between assortment size and category performance. On theone hand, a broader assortment is associated with higher sales(e.g., Borle et al. 2005;Dhar, Hoch, andKumar 2001) because alarge assortment makes it more likely for consumers to find aproduct that matches their needs (Boatwright and Nunes 2001).

On the other hand, reducing assortment size by removing low-selling products has no or positive impact on category sales insome settings (e.g., Broniarczyk, Hoyer, and McAlister 1998;Dreze, Hoch, and Purk 1994). This is because consumer choiceis affected by consumers’ perception of variety, which isdetermined not only by the assortment size but also by otherfactors, such as the availability of consumers’ favorite products(Broniarczyk, Hoyer, and McAlister 1998) and the number ofbrands in the category (Briesch, Chintagunta, and Fox 2009).Consequently, retailers struggle in aligning their assortmentswith consumer needs.

Even for a fixed assortment size, finding the right assort-ment composition is nontrivial because products withina category can cannibalize one another’s sales. The complexityof assortment decisions has been tackled by attribute-basedmodels that parsimoniously capture the interactions amongmanyproducts. Fader andHardie (1996) have developed a consumerchoice model that characterizes each product by its attributes(e.g., brand, size, color). Subsequent research has shown thatproductswith similar attributes (e.g., same size) aremore likelyto compete for demand within a category (Rooderkerk, VanHeerde, and Bijmolt 2011). Consequently, attribute-basedmodels have been used to make assortment decisions (e.g.,Rooderkerk, Van Heerde, and Bijmolt 2013).

Although the assortment research in the context of categorymanagement has focused on the retailer’s assortment decisions,several articles have studied assortment decisions in the contextof CC. For instance, Kurtulus and Nakkas (2011) show that CCleads to higher category sales because the captain’s consumerinsights enable the retailer to offer an assortment that is betteraligned with consumer needs. Furthermore, the captain canincrease its own sales through excluding some competingbrands from the category so that consumers switch to thecaptain’s products (Kurtulus and Nakkas, 2011; Kurtulus et al.2014). While CC benefits the retailer and the captain, it mayhurt or benefit competing manufacturers. On the one hand, CCcan lead to a decline in the shelf space allocated to competingmanufacturers (Kurtulus and Toktay 2011) or the captain mayrecommend removing a competitor’s product from the assort-ment (Kurtulus et al. 2014). On the other hand, CC is beneficialfor competing manufacturers’ products introduced to theassortment on a CC implementation (Kurtulus and Nakkas2011; Kurtulus et al. 2014). Overall, the CC literaturesuggests that both the captain and the retailer benefit,whereas the competing manufacturers may benefit or sufferafter CC is implemented.

Merchandising under category management. Althoughour data set does not have any variables associated withmerchandising, we briefly discuss merchandising efforts in thecontext of CC to provide a more comprehensive theoreticalbackground for our study. In practice, the captain’s merchan-dising efforts focus mainly on providing the retailer withdemand-enhancing services, such as shelf design, shelf-spaceallocation, and design of endcap displays (Subramanian et al.2010, p. 1741). The literature has provided support for thebenefits of such demand-enhancing services. For instance, aproduct’s sales can improve by increasing the number of shelffacings allocated to that product (e.g., Chandon et al. 2009) and

Implications of Category Captainship / 129

changing its shelf placement (e.g., Atalay, Bodur, andRasolofoarison 2012; Dreze, Hoch, and Purk 1994). More-over, taking into the account the decision sequence con-sumers follow to narrow down options (e.g., first flavor,then size) in organizing shelf displays (Nowlis, Dhar, andSimonson 2010) and designing shelves and aisles to increasethe proximity of complementary products can improvecategory performance (e.g., Bezawada et al. 2009; Dreze,Hoch, and Purk 1994).

The CC literature has also considered the benefits ofdemand-enhancing efforts. In particular, Subramanian et al.(2010) focus on the impact of the captain’s demand-enhancingservices on category stakeholders. They suggest that suchservices benefit the retailer by improving category performancebutmay benefit or hurt the competingmanufacturers dependingon whether they increase overall category demand or shiftdemand from one brand to another. Similarly, Kurtulus,Nakkas, and Ulku (2014) consider CC in a context in which thetotal category demand is a function of the effort that the retaileror the captain exerts into demand-enhancing services. In theirsetting, the competing manufacturers benefit (suffer) from thecaptain’s merchandising efforts if the captain keeps themin (removes them from) the assortment. In summary, theCC literature has documented that CC can improve cat-egory performance through the captain’s merchandisingefforts, and such efforts might benefit or hurt competingmanufacturers.

Conceptual Framework

Our theoretical discussion reveals that the key drivers of cat-egory decisions in our context are (1) category objective,(2) private label presence, (3) collaboration with a captain,(4) cross-price and cross-assortment effects, and (5) attribute-based substitution among products. Drawing on these drivers,we have developed a conceptual framework (see Figure 1).

In light of the main objective of the CC implementationwe study (i.e., improving category sales), our framework linksCC to sales. This is aligned with practice, in which assessingcategory performance through sales (rather than profitability) iscommon in theCC context.1 Although profit maximizationmaybe a better goal for the retailer in some categories, asking thecaptain to improve profitability would require the retailer toshare the competing manufacturers’ proprietary informa-tion (e.g., wholesale prices) with the captain. However,legal authorities discourage retailers from sharing one man-ufacturer’s proprietary information with another (Desrochers,Gundlach, and Foer 2003; Federal Trade Commission 2001).As a result, many retailers define category objectives in termsof sales in CC implementations (Kurtulus, Nakkas, and Ulku2014; Kurtulus et al. 2014).

Our framework posits that the captain might influence salesby making pricing, assortment, and/or merchandising recom-mendations. Moreover, manufacturer type (i.e., private label,captain, and competing manufacturers) and product attributesmight moderate these recommendations and their impact onsales. In particular, manufacturer type determines the extent towhich a manufacturer has control over category decisions, andproduct attributes determine substitution patterns within acategory.

Impact of CC on retail prices. Our theoretical discussionsuggests that implementing CC creates an alliance between theretailer and the captain, which in turn enables the retailer tolower retail prices for the captain’s products. It also reveals thatthe retailer may choose to lower retail prices for private labelproducts to boost private label sales. Combining theseobservations with the notion that the CC implementation westudy attempts to increase category sales, we expect that CCwill reduce retail prices for the captain’s and private labelproducts.

Two opposing forces affect retail prices for the competingmanufacturer’s products. On the one hand, the competingmanufacturers may respond to the captain’s and private label’sprice reductions by reducing their wholesale prices, which mayallow the retailer to lower those products’ retail prices. On theother hand, product attributes and cross-price effects mayinfluence pricing decisions. In particular, the competing man-ufacturers’ products that are in direct competition with thecaptain’s products and/or private label products may be pricedhigh to ensure that the captain’s products and private labelproducts are better positioned against their close competitors.Although we do not have data on wholesale prices, these twoforces suggest that the retail prices for the competing manu-facturers’may increase or decrease after the CC implementation.

Impact of CC on assortment. Our theoretical discussionsuggests that the captain’s products may receive a preferentialtreatment after CC. Furthermore, the captain’s attempt toincrease private label sales may create a similar preferentialtreatment for the private label. Drawing on these observations,we expect that the CC implementation will increase theassortment presence of the captain’s products and privatelabel products. Our theoretical discussion also suggests that thecompeting manufacturers’ products may experience an increaseor a decline in their assortment presence. From the findingsrelated to the captain’s opportunistic behavior, we anticipate thatthe competingmanufacturers’ products that are close substitutesto the captain’s products may experience a decline in theirassortment presence. In contrast, the competing manufacturers’products that are not in direct competition with the captain’sproducts may experience an increase in their assortmentpresence so that category sales are increased. Because theretailer may protect its private label, a product’s similarityto the private label may also negatively influence its assortmentpresence.

Impact of CC on sales. Our conceptual frameworkdepicted in Figure 1 suggests that CCmay influence a product’ssales through three mechanisms: pricing, assortment, andmerchandising. First, the impact of pricing on a product’s sales

1For instance, 4 of the 11 performance criteria Progressive Grocerused to identify successful category captains in its 2016 categorycaptains of the year article emphasize sales, whereas none of thecriteria emphasize profitability (Progressive Grocer 2016, p. 42).Indeed, more than half of the 62 CC examples provided in the articlereport implementation results based on sales, whereas only 4 examplesmention profitability.


will be determined by a combination of own- and cross-priceeffects.We expect the own-price elasticity to be negative and thecross-price elasticity to be positive. Second, the impact ofassortment changes on a product’s sales will be determined byown- and cross-assortment effects. We expect the own-assortment elasticity to be positive and the cross-assortmentelasticity to be negative. Third, the impact of CC might gobeyond pricing and assortment changes because the captainmay also make merchandising recommendations. In line withour theoretical discussion, we expect that the captain’s rec-ommendations on pricing, assortment, and merchandising (ifprovided) will each have a positive impact on the sales of aprivate label product or a product that belongs to the captain.However, the impact of those recommendations on a compet-ing manufacturer’s product might be positive or negative,depending onwhether that product is in close competition withthe captain’s products or private label products, as determinedby the products’ similarity.

Research SettingImplementation Summary

The retailer we study is a grocery chain that serves multiplestateswithin one of themajor geographical regions in theUnitedStates. It is one of the largest grocery retailers in its region interms of market share and sales volume. The retailer has a

reputation for offering high-quality products. The category westudy is a mature, shelf-stable food category (i.e., a center-storegrocery product). As per the retailer’s request, the soon-to-becaptain manufacturer provided the retailer with a strengths,weaknesses, opportunities, and threats (SWOT) analysis of thecategory.

The main findings of the captain’s SWOT analysis, whichwe obtained from a presentation given by the captain to theretailer prior to the CC implementation, were as follows.The main strengths were the retailer’s strong reputation as theprovider of high-quality grocery products, its local marketfamiliarity, and its loyal customer base. The main weaknesswas the retailer’s lack of category knowledge, leading to poorassortment and pricing decisions. In particular, the captainhad identified that despite offering a similar assortment withrespect to its competitors, the retailer’s average unit price of$.72 in this category was $.07 higher than the rest of market.Moreover, the average unit price for the retailer’s private labelwas $.66, which was $.13 higher than its competitors’ privatelabels. The biggest opportunities were to leverage the retailer’sreputation and the captain’s category knowledge to improveprivate label performance and increase category revenues. Themain threat for the retailer was to lose furthermarket share to itscompetitors. In light of the captain’s SWOT analysis, theretailer decided to receive additional help through a formal CCimplementation. Under the captain’s guidance, assortment andpricing recommendations were generated and implemented.

FIGURE 1Conceptual Framework andDirectional Expectations for the Impact of theCConSales for a Particular Product

Product AttributesManufacturer Type

• Category captain• Private label• Competitive set

Category

CaptainshipSales

Assortment• Own• Competitors’

Pricing• Own• Competitors’

Directional Expectations Manufacturer Type

Private Label Category Captain Other Manufacturers

Impact of CC on price − − +/−Impact of CC on

Impact of CC on sales assortment

++ +/−

++ +/−

Notes: Pricing and assortment boxes illustrate that theCC implementation affects pricing and assortment, which in turn affect sales through own and cross-elasticities. Because our data set does not include any variables on merchandising, we model merchandising as a direct link from CC to sales.Manufacturer type and product attributesmoderate the impact of CCon pricing, assortment, andmerchandising changes.Wediscuss the empiricaloperationalization of these three mechanisms in the “Model” section.


Data Description

Our data set spans 52 weeks and consists of weekly UniversalProduct Code (UPC)–level measures for all products in thecategory. In week 21, the retailer modified the category usingthe services of a captain, which is the second-largest manu-facturer in the category after the private label. Our data areaggregated to the firm level and include the following metricsfor product i in week t:

• Sales revenue, rit: The cumulative sales revenue generatedfrom all stores.

• Sales quantity, qit: The cumulative number of units (pack-ages) sold across all stores.

• Product distribution, dit: The percentage of stores that carry aparticular product, weighted by the relative size of thosestores. Formally, this measure of distribution is referred to asthe all commodity volume–weighted distribution, whichrefers to the total annual sales revenue of a given store.

• Product size, zi: Volume of product i in ounces.

The aggregated structure of our data set prevents us fromobserving store- and product-level promotions and discounts.However, having access to rit, qit, dit, and zi enables us tocompute the average per ounce price consumers paid forproduct i in week t. Specifically, when product i is carried inthe assortment in week t (i.e., when dit > 0), we define itsaverage volumetric price (i.e., price per ounce) as

pit =ritqitzi

.(1)

Moreover, our data set includes a base price for each product.When a product is not carried in the assortment in a particularweek (i.e., when dit = 0, qit = 0, and rit = 0), we use this baseprice (after converting it to a volumetric price) as pit.

Figure 2, Panel A, shows that weekly sales are higher duringthe postimplementation period. It also shows that there are threespikes in the weekly category sales in weeks 28, 32, and 47,which correspond to Thanksgiving, Christmas, and Easter,respectively. Thus, at least some of the increase in salesmight bedriven by seasonality and increased consumption in these threeweeks. Presumably, when the consumption of a food itemincreases, consumers are more likely to search the web forrecipeswith the itemof interest as an ingredient. Thus,wewouldexpect the search frequency to capture a portion of variation inweekly sales. Figure 2, Panel B, shows the relative searchfrequency of the search phrase “category name recipe” in logscale. We obtain search frequency data from Google Trends,which is a publicly available web page providing time seriesdata regarding the relative search frequency of a keyword orsearch phrase. We observe a strong positive correlation of .83betweenweekly sales andweb search patterns. Accordingly, weuse gt, which denotes the web search frequency for the categoryin week t, to control for seasonality. We also control for holidayeffects by including a dummy variable, ht, that assumes a valueof 1 if a national holiday occurs in week t.

There are ten manufacturers in our data set. Each manu-facturer owns one brand in the category. Accordingly, we usemanufacturer and brand interchangeably herein. Table 1 pro-vides an overview of these manufacturers offering a total of 110UPCs. The private label has the largest share both in terms of thenumber of UPCs and sales revenue. It offers low-price productswith an average volumetric price of $.07 during the pre-implementation period. The category captain is the secondlargest manufacturer. It offers premium products with anaverage volumetric price of $.11 during the preimplementationperiod. The remaining eight manufacturers offer a total of 44

FIGURE 2The Weekly Category Sales and Search Frequency for the Category in the Same Period

100

150

200

250

300

Cat

ego

ry S

ales

($

Th

ou

san

ds)

Week

Go

og

le S

earc

h F

req

uen

cy

0 10 20 30 40 50

Week

0 10 20 30 40 50

1

2

3

4

A: Sales B: Search Frequency


UPCs. Some competing manufacturers offer low-price prod-ucts, whereas others offer premium products. For instance, theaverage volumetric prices for manufacturers 3 and 4 during thepreimplementation period are $.08 and $.13, respectively.

The weekly average category sales are 37.24% higher afterCC. The increase in sales after the implementation is dis-proportionately higher for the captain’s products comparedwith the competing manufacturers’ products. For example,Table 1 shows that the total average weekly sales of the captainincreased from $18,587 to $40,076, whereas the third manu-facturer’s total average weekly sales decreased from $10,739 to$6,947.

The key decision unit in a categorymanagement initiative isthe subcategory or switching level, rather than the categoryitself. In this context, a subcategory is defined as a collection ofproducts presumed to have similar product characteristics. Forexample, in the sugar and sweeteners category, subcategorieswould include brown sugar, granulated sugar, powderedsugar, artificial sweeteners, and so on. Our data set consists ofnine subcategories. Larger subcategories typically have moremanufacturers. The private label and the captain have apresence in all subcategories.

Because a subcategory is more homogeneous than a cat-egory in terms of product characteristics, it is usually assumedthat the collection of products in a subcategory are perfect ornear-perfect substitutes (Kok and Fisher 2007). Figure 3illustrates that such a substitution pattern is unlikely to hold inour data set. For instance, Figure 3, Panel A, shows that mostproducts in subcategory 2 have a volume of ~15 oz., but ~8 oz.and ~11 oz. products are also available in the assortment.Similarly, Figure 3, Panel B, illustrates that some ~15 oz.products in subcategory 4 are relatively cheap, with volumetricprices around $.09 per ounce, whereas another ~15 oz. productin the same subcategory has a volumetric price of $.14 perounce. These examples indicate that substitution among similarproducts (in terms of volumetric price and size) can be higherthan substitution among dissimilar ones. Such a substitutionpattern is likely to affect assortment changes made during theCC initiative. Accordingly, we control for product similarity inour empirical analysis. We measure the similarity between

products i and j in the same subcategory s using the fol-lowing metric:

dij = 1 -

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi:5

�pi - pj

psmax - psmin

�2

+ :5

�zi - zj

zsmax - zsmin

�2s

.(2)

In Equation 2, zi is the package size, and pi ” ð1=20Þ�20t=1pit is

the average volumetric price prior to the CC implementation forproduct i. In addition, psmax and p

smin are the average volumetric

prices of the most expensive and the cheapest products insubcategory s, respectively. Similarly, zsmax and z

smin are the sizes

of the largest and the smallest products by volume in sub-category s, respectively. psmax - psmin and z

smax - zsmin allow us to

normalize the impact of volumetric price and size, respectively,so that dij takes values in the unit interval ½0, 1�. As such, twoproducts with the exact same size and average volumetric pricewould have a similarity score of 1.2

Figure 4, Panel A, demonstrates the change in the averagedistribution of every UPC in our data set after CC. We observethat a vast majority of products offered by the private label(manufacturer 1), the category captain (manufacturer 2), andmanufacturer 4 experience an increase in their distributions.Conversely, most of manufacturer 3’s products experience adecline in their distributions. Recall from Table 1 that theaverage weekly sales of manufacturers 1, 2, and 4 increase,whereas manufacturer 3 experiences a decline in sales after CC.These observations are consistent with our conceptual frame-work suggesting that the assortment changes made during the

TABLE 1An Overview of Each Manufacturer’s Weekly Sales Before and After CC Implementation

Firm # of UPCs

Weekly Sales Pre-CC ($) Weekly Sales Post-CC ($) % Change inAvg. SalesMean Median SD Mean Median SD

1 (PL) 38 40,625 37,134 11,127 50,894 45,920 12,360 25.28%2 (CC) 28 18,587 18,072 2,092 40,076 24,529 39,896 115.61%3 16 10,739 10,684 1,354 6,947 6,419 2,329 -35.31%4 13 8,668 7,341 5,218 12,317 8,347 9,492 42.10%5 3 4,975 4,632 677 5,333 5,132 918 7.21%6 4 2,112 2,088 414 1,684 1,419 735 -20.28%7 4 296 298 31 286 263 77 -3.61%8 1 229 230 40 252 243 43 10.03%9 1 54 53 11 81 73 47 50.52%10 2 0 0 0 552 538 290 —

All firms 110 86,285 82,385 12,438 118,421 107,082 42,864 37.24%

Notes: PL = private label.

2An alternative way to define product similarity is to incorporatesubcategory and brand into dij. For instance, dij can be defined asdij = 1 -

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi:25½ðpi - pjÞ/ðpsmax - psminÞ�2 + :25½ðzi - zjÞ/ðzsmax - zsminÞ�2 + :25IMij + :25ISij

q,

where IMij is an indicator variable that takes a value of 0 if i and j belongto the same manufacturer, and 1 otherwise, and ISij is an indicatorvariable that takes a value of 0 if i and j are in the same subcategory,and 1 otherwise. Incorporating brand and subcategory effects into oursimilarity metric does not change our findings. Thus, we use Equation2 in our analysis because of its parsimony. The Web Appendixprovides estimation results for alternative model specifications thatincorporate subcategory and brand into our similarity metric.


implementation will benefit the private label and the captain’sproducts and that such changes may explain changes in sales.

Figure 4, Panel B, shows the change in the averagevolumetric price of every UPC in our data set. The averagevolumetric price declines for most products after CC. Thisobservation is aligned with the theoretical predictions in theliterature regarding the impact of CC on prices (Kurtulus andToktay 2011; Nijs, Misra, and Hansen 2014). Consistent withour conceptual framework, declining prices could be anotherreason for the increase in sales after CC.

ModelWe study the impact of CC on sales by building an empiricalmodel based on our conceptional framework illustrated inFigure 1. Specifically, we estimate a simultaneous system ofequations in which we allow CC to exert both a direct andindirect (through price and assortment) influence on sales. Wespecify each equation in log-log form, which enables us tointerpret a model coefficient as the expected proportionalchange in the dependent variable per proportional change in theindependent variable (Gelman and Hill 2006). Furthermore,log-log model specification leads to mathematically tractableexpressions regarding the impact of the CC implementationon prices, assortment, and sales. Logarithmic demand modelshave been used extensively in both marketing and economicapplications (Hanssens, Parsons, and Schultz 2003; Lilien,Kotler, and Moorthy 1992). Because our data follow apanel structure, we fit a hierarchical model that allows forheterogeneity in parameters for each individual UPC. We

begin by specifying the direct impact of CC and otherdemand-generating factors on volumetric sales using thefollowing specification:

logðvitÞ = b0i + b1i logðpitÞ + b2ipit + b3i logðditÞ + b4idit+ b5i logðgtÞ + b6iht + b7it + b8iIðt > tÞ + esit,

(3)

where

• vit are volumetric sales (i.e., qit · zi) for UPC i in week t,• pit is an index of competitive prices for UPC i in week t. We

construct this index by computing a weighted average log-arithmic volumetric price of all other products in UPC i’ssubcategory. Formally, pit = �j2A sðiÞnfigwij logðpjtÞ, whereA sðiÞ denotes the set of products that belong to product i’ssubcategory, sðiÞ, and wij = dij=ð�k2A sðiÞnfigdikÞ is the weightof product j, which captures the notion that products withsimilar attributes (i.e., high wij values) are closer competitors.

• dit is an index of competitive distribution for UPC i in week t.We measure the competitive distribution index as a weightedaverage logarithmic distribution of all other products in UPCi’s subcategory. Formally, dit = �j2A sðiÞnfigwij logðdjtÞ.3

FIGURE 3Product Positions with Respect to the Average Volumetric Price and Product Size in Subcategories 2 and 4

Ave

rage

Vol

umet

ric P

rice

($)

10 15 20 25 30

Private labelCategory captainManufacturer 3Manufacturer 5

Size (Oz.)

A: Subcategory 2 B: Subcategory 4

Size (Oz.)8 10 12 14

.06

.08

.10

.12

.14

.16

.18

Ave

rage

Vol

umet

ric P

rice

($)

.08

.10

.12

.14

.13

.11

.09

Private labelCategory captainManufacturer 3Manufacturer 4

Manufacturer 6Manufacturer 7Manufacturer 8

3We calculate cross-price and cross-distribution at the sub-category level because a subcategory is more homogeneous than acategory in terms of product characteristics. Alternatively, one cancalculate cross-price and cross-distribution using weights (i.e., wijvalues) derived from similarity scores calculated at the categorylevel (e.g., the formula presented in footnote 2). Calculating cross-price and cross-distribution at the category level does not change ourfindings. The Web Appendix provides estimation results for analternative model specification with category-level cross-price andcross-distribution variables.


• t is a linear indicator of time that enables us to model trend inthe time series.

• Iðt > tÞ is an indicator variable that is equal to 0 in the weekspreceding CC and is equal to 1 in the weeks following. Thevariable t denotes the week of implementation.

Finally, as defined in the previous section, gt and ht denotethe search frequency and holiday dummies, respectively.Equation 3 allows us to understand the direct impact of CC onsales performance through b8i. Because we define merchan-dising as all demand-enhancing actions excluding pricing andassortment, we interpret b8i as the impact of merchandising onproduct i’s volumetric sales.4 We capture the indirect impactof CC on sales through pricing and assortment by building asystem of equations as follows:

logðpitÞ = a0i + a1i logðgtÞ + a2iht + a3iIðt > tÞ + epit, and(4)

logðditÞ = g0i + g1i logðgtÞ + g2iht + g3iIðt > tÞ + edit.(5)

In Equations 4 and 5, coefficients fa1i,a2ig and fg1i, g2igcontrol for seasonality and holiday effects, whereas coefficientsa3i and g3i capture the impact of CC on UPC i’s price anddistribution, respectively.

We allow for correlation in the cross-equation error terms byassuming a multivariate normal structure for the joint dis-tribution of fesi , epi , edi g ~ Nð0,SiÞ. By doing so, we control forthe potential of unobserved demand shocks that could affect ourability to infer the true impact of CC on sales performance. Si

is a 3 · 3 covariance matrix that captures the degree to whichunobserved shocks to our demand system jointly influenceprice, distribution, and sales.

Because our model is estimated using hierarchical Bayesianmethods, we complete the hierarchy by specifying a distributionof heterogeneity over the complete vector of regression coef-ficients (Rossi, Allenby, and McCulloch 2005):

fbi,ai, g ig ~ NðDwi,WÞ,(6)

where wi are UPC-specific characteristics that can moderate therelationship between CC and sales, price, and distribution. D isan estimated matrix of coefficients that characterize this rela-tionship, and W is a full covariance matrix that captures cross-UPC correlation in the estimated coefficients. Included in wi arethe following variables:

FIGURE 4Change in Average Weekly Distribution and Volumetric Price for Each Product

Products

−.6

Manufacturer 1 (PL)

Ch

ang

e in

Ave

rag

e V

olu

met

ric

Pri

ce (

Cen

ts)

−6

6

4

−2

2

0

−4

Ch

ang

e in

Ave

rag

e D

istr

ibu

tio

n

−.2

.2

.6

A: Change in Distribution

B: Change in Volumetric Price

Manufacturer 2 (CC) Manufacturer 3 Manufacturer 4

Manufacturer 4Manufacturer 1 (PL) Manufacturer 2 (CC) Manufacturer 3

Products

Notes: PL = private label. Each bar represents a UPC. The vertical dashed lines separate different manufacturers. The manufacturers are displayed in adecreasing order from left to right in line with the preimplementation averageweekly sales revenues. The change in the averageweekly distributionfor UPC i is computed as Ddit = ð1=32Þ�52

t=21dit - ð1=20Þ�20t=1dit. The change in the average weekly volumetric price is computed similarly.

4While our data set does not have any variables that enable us toquantify the impact of specific merchandising efforts, it is plausiblethat the CC implementation may have led to some merchandisingchanges (e.g., new shelf displays, reallocation of shelf space amongproducts; Subramanian et al. 2010). b8i captures the aggregateimpact of such merchandising efforts.


• Iði2CCÞ is an indicator variable that is equal to 1 if UPC ibelongs to the category captain. The reference level for thisvariable is the private label, so the resulting coefficient shouldbe interpreted as one deviation away from the average privatelabel effect.

• Iði2CSÞ is an indicator variable that is equal to 1 if UPC ibelongs to a manufacturer that is in the competitive set, CS.The competitive set includes all manufacturers in the cat-egory, except the private label manufacturer and the captain.This variable’s effect should also be interpreted with respectto the private label.

• SimPLi is a continuous variable that measures the similarity

between UPC i and the private label products in UPC i’ssubcategory. Let A PL

s , A CCs , and A CS

s denote the subset ofproducts offered by the private label, captain, and competingmanufacturers in subcategory s, respectively. For UPC i, wecalculate SimPL

i as

SimPLi = max

j2A PLs

dij.(7)

In words, SimPLi is the maximum similarity between UPC

i and the private label products in its subcategory. A highSimPL

i value indicates that UPC i has a close substituteoffered by the private label manufacturer. By definition,SimPL

i = 1 for private label products. As we discussed in the“Theoretical Background” section, similar attribute-basedsubstitution models are used in the choice modeling andassortment literature streams (e.g., Hoch, Bradlow, andWansink 1999; Rooderkerk, Van Heerde, and Bijmolt2011, 2013).

• SimCCi is a continuous variable that measures the sim-

ilarity between UPC i and the category captain’s productsin UPC i’s subcategory. We compute this variable asSimCC

i = maxj2A CCsdij. By definition, SimCC

i = 1 for thecaptain’s products.5

We include Iði2CCÞ, Iði2CSÞ, SimPLi , and SimCC

i in ourupper-level model as covariates because the impact of theCC implementation may differ by manufacturer type as wellas whether a product is in direct competition with a productoffered by the private label manufacturer or the captain. Weestimate the parameters of the hierarchical model specifiedby Equations 3, 4, 5, and 6 using standard Bayesian methods(Rossi, Allenby, and McCulloch 2005). Specifically, weuse a hybrid sampler where the collection of regressionparameters for each UPC, fbi,ai, g ig, are drawn using theMetropolis–Hastings algorithm. Conditional on a realiza-tion of the regression parameters for all UPCs, we use aGibbs draw for both Si and the parameters in the upper level,D andW. We ran the sampler for 100,000 iterations and usedthe final 50,000 draws for inference.

After estimating model parameters, we quantify the effectof the CC implementation by measuring its impact on cat-egory sales revenues. Using volumetric sales and volumetricprices, the sales revenue for UPC i in week t can be expressedas rit = vitpit. In the log form, we have logðritÞ = logðvitÞ +logðpitÞ. Using the expression for logðvitÞ specified in Equation3, we obtain

logðritÞ = b0i + ð1 + b1iÞlogðpitÞ + b2ipit + b3i logðditÞ + b4idit+ b5i logðgtÞ + b6iht + b7it + b8iIðt > tÞ + esit.

(8)

Accordingly, we measure the impact of the CC imple-mentation on UPC i’s sales revenues by taking a partialderivative of Equation 8 with respect to I, the CC indicatorvariable. That is,

¶logðritÞ¶I

= ð1 + b1iÞ¶logðpitÞ

¶I+ b2i �

j2A sðiÞnfigwij

¶logðpjtÞ¶I

+ b3i¶logðditÞ

¶I+ b4i �

j2A sðiÞnfigwij

¶log�djt�

¶I+ b8i.

(9)

Because CC influences prices and distribution, the right-handside of this equation has partial derivatives of the price, cross-price, distribution, and cross-distribution variables with respectto I. Differentiating the price and the distribution equationsof our empirical model, 4 and 5, with respect to I gives¶logðpitÞ=¶I = a3i and ¶logðditÞ=¶I = g3i, respectively. There-fore, we can rewrite Equation 9 as

¶logðritÞ¶I

= ð1 + b1iÞa3i + b2i �j2A sðiÞnfig

wija3j + b3ig3i

+ b4i �j2A sðiÞnfig

wijg3j + b8i$

(10)

In Equation 10, ð1 + b1iÞa3i + b2i�j2A sðiÞnfigwija3j capturesthe total impact of the changes in UPC i’s price and the com-petitors’ prices, b3ig3i + b4i�j2A sðiÞnfigwijg3j captures the impactof the changes in UPC i’s and its competitors’ distribution, andb8i captures the impact of merchandising efforts on UPC i’ssales. In the next section, we use Equation 10 to measure anddecompose the impact of CC on the entire category as well asthe individual manufacturers.

ResultsTable 2 shows the coefficient estimates for the sales, price, anddistribution equations specified in Equations 3, 4, and 5, re-spectively. We report the posterior mean and 95% interval foreach coefficient. The CC indicator coefficient in the salesequation indicates thatmerchandising efforts increase categorysales by 8.6%, which is significant at p < .05. (Hereinafter, wereport statistical significance at p < .05.) We also find that thesearch frequency variable is positively correlated with sales.The holiday indicator variable is insignificant, which impliesthat the search frequency variable captures not only seasonalitybut also the sales spikes observed during holidays. Consistentwith the well-known impact of pricing, a product’s salesdecrease in its own price and increase in its competitors’ prices.

5An alternative way to measure a product’s similarity to privatelabel or CC is to calculate an average similarity score (e.g.,ð1=��ACC

s

��Þ�j2A CCsdij). Our findings remain qualitatively unchanged

when we use the average similarity (instead of the maximumsimilarity) scores as explanatory variables. We opt in for maximumsimilarity scores because they better capture close competitionbetween two UPCs with similar attributes, whereas the averagesimilarity scores are less informative about close competitionbetween similar products.


Sales increase in own-distribution, but we do not find a directrelationship between competitive distribution and sales.6Finally, there is no time trend in sales.

The CC indicator coefficient in the price equation indicatesthat CC leads to a 9.2% decline in volumetric prices, which isstatistically significant. Both the holiday indicator and searchfrequency variables are negative and significant, indicating thatthe sales increases during the holiday weeks are in part due tothe steep price reductions in those weeks. The CC indicatorcoefficient in the distribution equation is insignificant. Thatis, we do not observe a systematic increase or decline in theoverall distribution offered by the retailer in this category, onaverage.

After estimating the model coefficients, we use Equation 10to quantify the impact of CC on different stakeholders.Specifically, we first use the posterior distributions of thecoefficients that appear in Equation 10 to obtain the impactof merchandising, assortment, and pricing changes on UPCi’s sales revenues. Then, we calculate the impact of CC on aproduct subset (e.g., the captain’s products) by taking aweighted average of the impacts on each UPC in thatsubset, where the weight for each UPC equals its pre-implementation market share within that subset. Table 3shows the impact of CC on the entire category as well as themanufacturers.

Impact of the CC Implementation on Sales

Table 3 shows that the increase in category sales that can beassociated with CC is 19.1%, which is statistically significant.The pricing changes boost category sales by 7.6%. Theassortment changes increase sales by 2.9%. Finally, mer-chandising efforts lead to a 8.6% increase in category sales. Theimpacts of pricing, assortment, and merchandising on categorysales are all statistically significant. Table 3 also shows that CCleads to a 13.9% increase in private label sales and a 42.4%increase in the captain’s sales. The impacts of merchandising,pricing, and assortment are all significant for both the privatelabel and the captain’s products.

Our analysis shows that some manufacturers benefit fromCC, whereas others suffer as manifested by declining sales. Forinstance, Table 3 shows that CC decreases manufacturer 39ssales by 18.2%. This decrease is driven by the decline in thepresence of manufacturer 3’s products in the assortment.Manufacturer 6 experiences a sales decline as well. Thedecrease in manufacturer 6’s sales due to CC is 33.4%. A largeportion of this decrease (i.e., 15.6%) is driven by the declineof manufacturer 6’s presence in the assortment. Contrary tomanufacturers 3 and 6, manufacturer 4 benefits because CCincreases its sales by 37.1%.Most of this increase (i.e., 19:3%)is due to assortment changes. The impact of CC on manu-facturer 5’s sales is statistically insignificant. The remainingfour manufacturers have relatively small sales, which makes itdifficult to provide reliable estimates at a manufacturer level.However, the total impact of CC on these fourmanufacturers ispositive, as manifested by a 13.1% increase in their total salesthat can be attributed to CC.

There are two main observations that emerge from ana-lyzing the changes in sales by manufacturer. First, both theprivate label and the category captain’s products benefit fromCC. Second, some competing manufacturers also benefit fromCC, whereas others suffer as manifested by their decliningpresence in the assortment and lower sales. While the firstobservation is aligned with the theoretical predictions in theliterature, the second observation raises a follow-up question:

TABLE 2Coefficient Estimates of the Sales, Price, and Distribution Equations

Sales Equation Price Equation Distribution Equation

Mean LB UB Mean LB UB Mean LB UB

Intercept 6.603 6.510 6.700 -2.405 -2.504 -2.304 .610 .512 .702Own price -1.393 -1.500 -1.300Cross-price .145 .053 .230Own distribution 1.548 1.461 1.642Cross-distribution .024 -.065 .111Search frequency .193 .105 .277 -.138 -.216 -.061 .021 -.092 .093Holiday .088 -.008 .177 -.122 -.199 -.040 .017 -.061 .099Linear trend -.005 -.081 .072CC indicator .086 .001 .166 -.092 -.169 -.004 .020 -.062 .099# of UPCs 110 110 110Sample size 5,720 5,720 5,720R2 .905 .317 .357

Notes:We report coefficient estimates for Equations 3, 4, and 5. LB andUB are the 2.5%and 97.5%quantiles of the estimated parameter’s posteriordistribution, respectively. Statistically significant values at the 95% level are highlighted in bold.

6Our UPC-level analysis (not reported herein because of spacelimitations) revealed that the insignificance of the cross-distributionvariable at the aggregate level is driven by two factors. First, a smallnumber of high-demand UPCs are insensitive to the presence ofsubstitute products. Second, own- and cross-distribution variablesmove in the opposite directions for some UPCs. For instance, if aproduct’s distribution score declines, its close competitors’ dis-tribution scores increase, creating a colinear relationship betweenown- and cross-distribution variables. For these two reasons, thecross-distribution variable is insignificant at the aggregate level.Nevertheless, we show that demand substitution plays a crucial rolein shaping the postimplementation assortment.


Why do some competing manufacturers benefit from CC,whereas others experience adverse consequences? We addressthis question next.

Competitive Implications of Category Captainship

In this subsection, we focus on the products offered by thecompeting manufacturers and establish a link between productattributes and assortment changes.We begin our analysis with amotivating example. Table 4 provides a summary of productattributes in subcategory 9, which is the smallest subcategoryin our data set. Comparing the pre- and postimplementationaverage distributions shows that product 2 offered by manu-facturer 3 is removed from the category after CC.7 A closerexamination of product attributes reveals that product 1, whichis a private label product, and product 2 have the same size andaverage volumetric price. Thus, one potential explanation forthe removal of product 2 from the assortment is an attempt toincrease private label sales by excluding the private label’sclose competitors.

In light of the aforementioned example from subcategory 9,we conjecture that a competing manufacturers’ product is morelikely to experience a decline in its distribution if there is a closesubstitute to that product offered by the private label manu-facturer. Similarly, a product’s similarity with the categorycaptain’s products may also have a negative impact on itsassortment presence after CC. The hierarchical structure of ourempirical model enables us to test our conjectures regarding thepotential negative impact of a product’s similarity to the privatelabel and/or the captain’s products. Table 5 reports the upper-level model coefficients for the CC indicator variables in thesales, price, and distribution equations (i.e., Equations 3–5).Because manufacturers 3 and 6 suffer and manufacturer 4benefits from the assortment changes, we focus on the CCindicator coefficient of the distribution equation, g3i. For aparticular UPC i offered by a competing manufacturer, our

hierarchical model specification implies that the estimatedvalue of g3i can be written as

¶logðditÞ¶I

= bg3i = :020 + :002Iði2CCÞ - :087Iði2CSÞ- :249SimPL

i + :255SimCCi , and

(11)

= - :067 - :249SimPLi + :255SimCC

i ,(12)

where the second line follows because Iði2CCÞ = 0 andIði2CSÞ = 1 for a product offered by a competing manu-facturer. Equation 12 captures the estimated change in theassortment presence of a product offered by a competingmanufacturer as a function of its similarity to the private labeland the captain’s products. The intercept, -.067, is statisti-cally insignificant. The coefficient of SimPL

i , which measuresthe similarity between UPC i and the private label products inthe same subcategory, is negative and significant. That is, aproduct is more likely to experience a decline in its assort-ment presence if there is a similar private label product inthe assortment. Conversely, the coefficient of SimCC

i , whichmeasures the similarity between UPC i and the captain’sproducts in the same subcategory, is positive and significant.This finding implies that the competing manufacturers’products that are similar to the captain’s products aremore likelyto increase their presence in the assortment during the post-implementation period.

These findings are consistent with the objectives of the CCimplementation we study. In particular, decreasing the assort-ment presence of manufacturers 3 and 6, which are in directcompetition with private label because of their low prices,leaves the private label as the most affordable option in eachsubcategory. Consequently, private label performance im-proves. Furthermore, increasing the assortment presence ofmanufacturer 4, which offers premium products, enables theretailer to enrich its assortment and thereby increase overallcategory sales. While the protection of private label is in linewith the existing literature (e.g., Chintagunta 2002), theincreased presence of manufacturer 4 contrasts the liter-ature’s predictions on the negative impact of CC on thecompeting manufacturers.

TABLE 3The Decomposition of the Impact of Category Captainship by Manufacturer

Impact ofMerchandising Impact of Price Impact of Assortment Total Impact

Manufacturer Mean LB UB Mean LB UB Mean LB UB Mean LB UB

1 (PL) .045 .035 .052 .026 .022 .030 .068 .052 .090 .139 .127 .1552 (CC) .214 .197 .232 .092 .085 .097 .118 .092 .142 .424 .396 .4523 -.004 -.026 .029 -.013 -.029 .003 2.165 -.187 -.136 2.182 -.206 -.1554 .113 .089 .143 .065 .046 .080 .193 .124 .235 .371 .311 .4155 -.014 -.055 .059 .037 .025 .053 .023 -.022 .085 .045 -.013 .1716 2.187 -.220 -.155 .008 -.020 .029 2.156 -.224 -.080 2.334 -.413 -.258All other .109 .087 .145 .009 -.003 .029 .013 -.026 .065 .131 .101 .185Entire category .086 .001 .166 .076 .066 .088 .029 .003 .043 .191 .177 .219

Notes. PL = private label. LB and UB are the 2.5% and 97.5% quantiles of the estimated impact, respectively. Statistically significant values at the95% level are highlighted in bold.

7The average distribution for product 2 during the post-implementation period is slightly above zero because it took theretailer a few weeks to completely remove this product from theassortment.


What would have happened if the retailer had not protectedits private label?8 We address this question by revisitingEquation 11. This equation suggests that the retailer may beprotecting its private label by reducing the assortment presenceof the competing manufacturers’ products that are in closecompetition with private label. Thus, we conjecture that suchproducts would have had a greater assortment presence in theabsence of private label protection. Increasing the assortmentpresence of those productswould have influenced category salesthrough own- and cross-distribution effects.

We operationalize this conjecture by considering a scenarioin which the coefficient for SimPL

i equals zero instead ofits estimated value, -.249. Setting this coefficient to zeroimplies that a product’s similarity to private label does not leadto a decline in its assortment presence after CC. For instance,revisiting the example presented in Table 4, product 2 insubcategory 9 has SimPL

i = 1 because product 1, which is aprivate label product, has the same size and average volumetricprice as product 2. Equation 11 suggests that this product’ssimilarity to private label is associated with a -:249 ·SimPL

i = - :249 decline in its assortment presence. Thus, in theabsence of private label protection, we conjecture that thisproduct’s postimplementation distribution would have beenhigher by .249. This change would have also increased thecross-distribution values for the other products in the samesubcategory. Applying the same logic to all products in thecategory enables us to calculate the expected own- and cross-distribution levels for each product in an alternative setting inwhich the private label is not protected. After calculating thenew own- and cross-distribution levels for each product, weplug them into Equation 3 to calculate the impact of CC on eachproduct’s sales in this alternative setting.

Figure 5 reports how switching from the original CCimplementation setting to an alternative setting in which theprivate label is not protected changes the impact of CC on theentire category as well as each manufacturer. In the alternativesetting, the overall category sales would have been 4.1 per-centage points higher than the increase we estimate under theoriginal setting. That is, the impact of CC on category sales is19.1% in the original setting as reported in Table 3, whereas itis 23.2% in the alternative setting. Private label sales decline by3.2 percentage points when private label is not protected.However, its main competitor, manufacturer 3, experiences a

13.8-percentage-point increase in its sales. The sales changesfor the remaining manufacturers in the category are mixedbecause changing SimPL

i affects not only a product’s owndistribution but also the distribution of the remaining productsin the category. It is important to note that although totalcategory sales are expected to increase under this alternativeregime, this finding does not necessarily imply that the retailermade a mistake in protecting the private label. Because we donot observe product margins, it is possible that the increase insales could be associated with a decrease in profitability.

Robustness Tests

We examined alternative model specifications to assess therobustness of our findings. First, it is plausible that fitting asingle coefficient for all competing manufacturers in the upper-level model may not fully capture the differences betweenmanufacturers. Thus, we expanded the upper-level modelvariables (i.e., wi) to include manufacturer-specific dummyvariables. Expanding the set of upper-level model variables didnot change our results. Second, wemoved the product similarityvariables SimPL

i and SimCCi from the upper-level model to the

lower model (i.e., Equations 3–5) to determine whether thischange affects our findings. This alternative model specificationalso revealed that the competing manufacturers that offer closesubstitutes to the private label suffer from CC, whereas thecompeting manufacturers that closely compete with the captainbenefit from CC. Our remaining findings (i.e., the impact of CCon various stakeholders) also remained unchanged. Finally, it isplausible that it took the retailer more than one week to fullyimplement CC. To ensure that implementation delays do notaffect parameter estimates, we reran our analysis after droppingweeks 21–24, which gives the retailer four weeks for the CCimplementation. Our results remained qualitatively similar inthis alternative model specification, in which weeks 25–52denote the postimplementation period.

ConclusionAlthough CC has been implemented by many retailers and hasbeen examined from a theoretical perspective, to the best of ourknowledge, our study is a first attempt to empirically examinethe outcomes of an actual CC implementation. Our researchdemonstrates how joint consideration of assortment and pricing,the presence of a private label, and product characteristics mayinfluence the outcomes of CC implementations. Nevertheless,further research is needed to test whether our findings generalize

TABLE 4Summary of Product Attributes in Subcategory 9

Product Manufacturer Size

Preimplementation Averages Postimplementation Averages

Vol. Price Dist. Mkt. Share Vol. Price Dist. Mkt. Share

1 Private label 14.5 oz. $.06 .73 51% $.05 .71 64%2 Manufacturer 3 14.5 oz. $.06 .28 20% $.09 .02 1%3 Category captain 14.5 oz. $.12 .21 20% $.11 .19 24%4 Private label 8 oz. $.11 .25 9% $.08 .29 11%

Notes: We report the average values of volumetric price, distribution, and market share during the pre- and postimplementation periods.

8We thank the area editor for suggesting that we explore thisquestion.


beyond the CC implementation we study. The rest of thissection summarizes our main findings and then discusses theirimplications for practitioners and researchers.We conclude thearticle by discussing the limitations of our study.

Three key findings emerge from the CC implementationwe study that should be of interest to both practitioners andresearchers. First, we find that CC improves category sales inour setting. Despite ample anecdotal evidence on the benefits of

CC, our research is the first to empirically document the impactof CC on category performance. We also decompose thisimpact into the effects of different levers used by categorycaptains in practice (i.e., pricing, assortment, and merchan-dising). Second, we examine the impact of CC on categorystakeholders and find that CC benefits the private label as wellas the captain.We expected an increase in private label sales inour setting because one of the opportunities identified in the

FIGURE 5Change in the Impact of CC on Sales When There Is a Switch from the Current Regime, Which Protects the

Private Label, to an Alternative Regime in Which There Is No Private Label Protection

15

10

5

0

Per

cen

tag

e P

oin

t C

han

ge

Manufacturer

Entire Category 1 (PL) 2 (CC) 3 4 5 6 All Others

Notes: PL = private label.

TABLE 5Upper-Level Model Estimates for the CC Indicator Variables of the Sales, Price, and Distribution Equations

Equation Variable Mean LB UB

Sales, b8i Intercept .086 .027 .140Category captain .181 .022 .327Competitive set (excluding private label) .000 -.231 .218Similarity to private label .004 -.573 .506Similarity to category captain .000 -.404 .395

Price, a3i Intercept -.138 -.193 -.083Category captain -.017 -.145 .117Competitive set (excluding private label) .065 -.146 .290Similarity to private label -.001 -.473 .492Similarity to category captain -.204 -.537 .130

Distribution, g3i Intercept .020 -.030 .074Category captain .002 -.151 .154Competitive set (excluding private label) -.087 -.248 .098Similarity to private label -.249 -.723 -.049Similarity to category captain .255 .050 .599

Notes:We report the upper-level coefficients for b8i, a3i, and g3i, which correspond to the coefficients of the CC indicator variables in the sales, price,and distribution equations (i.e., Equations 3–5), respectively. LB and UB are the 2.5% and 97.5% quantiles of the estimated parameter’sposterior distribution, respectively. Statistically significant values at the 95% level are highlighted in bold.


captain’s SWOT analysis was the potential to improve privatelabel performance. We also expected the increase in thecaptain’s sales because there should be some remuneration forthe effort expended to develop and implement a categorymanagement strategy. We further find that some competingmanufacturers benefit from CC, whereas others suffer. Thisfinding is consistent with the CC literature’s mixed predictionsregarding the impact of CC on the competing manufacturers.

Third, and most importantly, we examine how productattributes and the retailer’s desire to protect the private labelaffect category performance. Despite the existing literature’spredictions regarding the negative impact of the captain’sopportunistic behavior, we find that the competing manu-facturers that are in direct competition with the captain benefitfrom CC. However, the competing manufacturers that are indirect competitionwith the private label suffer fromCCbecauseof a decline in their assortment presence. Indeed,wefind that theretailer’s desire to protect its private label prevents the retailerfrom maximizing category sales. In particular, private labelsales would have decreased but the overall category sales wouldhave increased if the retailer had not protected its private label.We discuss the managerial implications of these findings next.

Managerial Implications

Our study leads to four main managerial insights. First, ouranalysis reveals that a significant portion of the increase incategory sales is due to lower prices. From the retailer’s per-spective, this finding is important as it illustrates potential perilsof using price as a lever. When initiating a CC relationshipwith a manufacturer, the retailer establishes a collection ofperformance targets that will be used to assess CC success. Ourconversationswith categorymanagers indicate that these targetsare typically defined in terms of sales rather than profit becauseretailers are reluctant to release margin information to thecaptain. Given that sales targets can be achieved by loweringprices in price-sensitive categories, retailers should providegreater specificity in terms of how these targets are to beachieved. Otherwise, the captain may rely heavily on pricing,which is relatively easy to implement but may be costly in thelong run because price reductions lower product margins andmay heighten consumer sensitivity to price. Improving theassortment, however, is likely more difficult and carries highinitial fixed costs, but it should be relatively costless in the longrun.

Second, by comparing the pre- and postimplementationsales shown inTable 1, onemight conclude that the captain usedits authority to adversely influence its competitors. A closerexamination of the category dynamics, however, reveals adifferent story. We find that the manufacturers that closelycompete with the captain experience an increase in theirassortment presence and sales. The positive impact of CC on thecaptain’s close competitors is in part driven by the retailer’sdesire to increase category sales by offering a more attractiveassortment. Recall that the retailer has a reputation for offeringhigh-quality products. Thus, the captain, which offers premiumproducts, would have had difficulty in justifying the removalof other premium brands (i.e., its close competitors) from theassortment.Moreover, given the retailer’s reputation, increasing

category sales would have been difficult in the absence ofpremium brands. These observations suggest that one way forretailers to minimize the risk of opportunistic behavior by thecaptain is to identify the captain’s close competitors andcarefully formulate performance objectives so that it is difficultfor the captain to justify the removal of those competitors fromthe assortment.

Third, our analysis sheds light on howprivate label presencemay influence category decisions and performance in the CCcontext. Consistent with the existing literature (e.g., AilawadiandHarlam 2004; Ailawadi, Pauwels, and Steenkamp 2008),our findings suggest that there is a delicate balance betweenpushing the private label and maximizing category per-formance. On the one hand, solely focusing on categoryperformance may lead to poor private label performance.On the other hand, putting too much emphasis on privatelabel may hurt national brand sales, which in turn negativelyaffects overall category performance. Thus, retailers shouldcarefully assess the advantages and drawbacks of privatelabel protection prior to working with a captain.

Finally, private label presence has implications for com-peting manufacturers. The existing literature suggests thatprivate label introduction, which provides affordable productoptions for consumers, can be beneficial for premiumbrands butcan harm second-tier (i.e., low-price) brands in a category (e.g.,Pauwels and Srinivasan 2004). Similarly, we find that theincreased presence of the private label after CC hurts thecompeting manufacturers that are in direct competition withprivate label. Thus, our findings suggest that offering productsthat are differentiated from private label in terms of price andproduct attributes (e.g., package size) can help the competingmanufacturers avoid being excluded from the category.

Implications for Researchers

We find that lower prices, a better assortment, and merchan-dising efforts all contribute to the increase in category sales inour setting. Given the joint importance of all three levers,conclusions drawn from theoretical models, which typicallyfocus on a single lever, should be treated with caution becausesuch models might underestimate the value of CC. Moreover,our findings regarding the negative impact of private label onthe competing manufacturers indicate that theoretical studiesshould not overlook the private label. In particular, the objectivefunctions used in modeling studies, such as revenue and profitmaximization, may not fully reflect the retailer’s desire toprotect its private label. Thus, similar to Chintagunta (2002),defining an objective function that balances private label andcategory performance may be more appropriate to analyze theadvantages and drawbacks of CC initiatives. Our approach ofmodeling demand substitution on the basis of product attributesalso differentiates our study from the existing CC literature.Although similar approaches have been frequently used in theassortment literature (e.g., Fader and Hardie 1996; Rooderkerk,Van Heerde, and Bijmolt 2013), the CC literature uses morestylized models (e.g., linear demandmodels) to capture demandsubstitution. Our findings suggest that an attribute-baseddemand substitution model may lead to more precise inferenceson the competitive implications of CC.


Limitations and Future Research

Our article constitutes an important contribution to the literatureon CC in large part because the uniqueness of our data setenables us to observe phenomena that have not been previouslyconsidered in the CC context. That said, our data set is alsolimited in a variety of ways, thus creating opportunities forfurther research.

First, our data set is limited to a single category and retailer.As such, our findings are idiosyncratic to the category, retailer,and the goal of the CC implementation. Further research isneeded to test the generalizability of our findings, especially tosettings where a retailer seeks to improve category profitability.Moreover, controlling for the retailer’s decision to initiate CCthrough a more comprehensive data set including multipleCC implementations and/or control categories may be useful.This is because our study may overstate the benefits of CCas the CC implementation decision in our context was inpart driven by the retailer’s relatively poor category per-formance. It is possible that switching to a CC regime couldhave a smaller impact in categories that are already per-forming well.

It is also possible that the relative importance of pricing toassortment as a driver of category sales could differ for maturecategories (such as ours) versus categories characterized byproduct innovation.We expect pricing to play a more importantrole in settings where the retailer attempts to maximize the sales

revenues of a mature category. In contrast, assortment mightplay a more important role in categories with frequent newproduct introductions because it might be relatively easy toboost category performance by introducing new products insuch categories. In line with these predictions, it would beuseful to replicate the models developed herein across manycategories and retailers. Suchwork could yield insights into theboundary conditions of the findings from this article and helpresearchers and practitioners understand when and under whatconditions CC is likely to be most effective.

Second, our data set does not contain information about theproducts’ marginal costs. A richer data set would enable us tostudy a variety of interesting phenomena, including retailerprofitability and consumer welfare implications of CC. Finally,because our data set is limited to a single retailer, it does notenable us to study cross-retailer effects. Prior research hassuggested that at least a portion of the sale increase experiencedby a manufacturer during a promotion is the result of cross-retailer cannibalization (Van Heerde, Leeflang, and Wittink2004). Given the right data, it would be worthwhile to considercross-retailer dynamics that result from CC implementations.This type of research would illuminate the long-term impli-cations of CC, including an understanding of the resultingcompetitive equilibrium. We hope that our study will pave theway for greater collaboration between retailers and researchers,facilitating further empirical research on CC.

REFERENCESACNielsen (2005), Consumer-Centric Category Management:

How to Increase Profits by Managing Categories Based onConsumer Needs. Hoboken, NJ: John Wiley & Sons.

Ailawadi, Kusum, Eric Bradlow, Michaela Draganska, VincentNijs, Robert Rooderkerk, K. Sudhir, et al. (2010), “EmpiricalModels of Manufacturer-Retailer Interaction: A Reviewand Agenda for Future Research,” Marketing Letters, 21 (3),273–85.

Ailawadi, Kusum, and Bari Harlam (2004), “An Empirical Analysisof the Determinants of Retail Margins: The Role Of Store-BrandShare,” Journal of Marketing, 68 (January), 147–65.

Ailawadi, Kusum, Koen Pauwels, and Jan-Benedict Steenkamp(2008), “Private-Label Use and Store Loyalty,” Journal ofMarketing, 72 (November), 19–30.

Atalay, Selin, Onur Bodur, and Dina Rasolofoarison (2012),“Shining in the Center: Central Gaze Cascade Effect on ProductChoice,” Journal of Consumer Research, 39 (4), 848–66.

Basuroy, Suman, Murali Mantrala, and Rockney G. Walters (2001),“The Impact of Category Management on Retailer Prices andPerformance: Theory and Evidence,” Journal of Marketing,65 (October), 16–32.

Besanko, David, Jean-Pierre Dube, and Sachin Gupta (2005),“Own-Brand and Cross-Brand Retail Pass-Through,”MarketingScience, 24 (1), 123–37.

Bezawada, Ram, S. Balachander, P.K. Kannan, and VenkateshShankar (2009), “Cross-Category Effects of Aisle and DisplayPlacements: A Spatial Modeling Approach and Insights,” Journalof Marketing, 73 (July), 99–117.

Blattberg, Robert C., and Edward J. Fox (1995), Category Man-agement: Getting Started, Guide 1. Washington, DC: ResearchDepartment, Food Marketing Institute.

Boatwright, Peter, and Joseph C. Nunes (2001), “ReducingAssortment: An Attribute-Based Approach,” Journal of Mar-keting, 65 (July), 50–63.

Borle, Sharad, Peter Boatwright, Joseph Kadane, Joseph C. Nunes,and Shmueli Galit (2005), “The Effect of Product AssortmentChanges on Customer Retention,” Marketing Science, 24 (4),616–22.

Briesch, Richard, Pradeep K. Chintagunta, and Edward J.Fox (2009), “How Does Assortment Affect Grocery StoreChoice?” Journal of Marketing Research, 46 (April),176–89.

Broniarczyk, Susan, Wayne Hoyer, and Leigh McAlister (1998),“Consumers’ Perceptions of the Assortment Offered in a GroceryCategory: The Impact of Item Reduction,” Journal of MarketingResearch, 35 (May), 166–76.

Carameli, Leo S., Jr. (2004), “The Anti-Competitive Effects andAntitrust Implications of Category Management and CategoryCaptains of Consumer Products,” Chicago-Kent Law Review,79, 1313–56.

Chandon, Pierre, Wesley Hutchinson, Eric T. Bradlow, and ScottYoung (2009), “Does In-Store Marketing Work? Effects of theNumber and Position of Shelf Facings on Brand Attention andEvaluation at the Point of Purchase,” Journal of Marketing,73 (November), 1–17.

Chintagunta, Pradeep K. (2002), “Investigating Category PricingBehavior at a Retail Chain,” Journal of Marketing Research,39 (May), 141–54.

Chintagunta, Pradeep K., Andre Bonfrer, and Inseong Song (2002),“Investigating the Effects of Store-Brand Introduction onRetailer Demand and Pricing Behavior,” Management Science,48 (10), 1242–67.


Corstjens, Marcel, and Rajiv Lal (2000), “Building Store LoyaltyThrough Store Brands,” Journal of Marketing Research,37 (August), 281–91.

Desrochers, Debra, Gregory Gundlach, and Albert Foer (2003),“Analysis of Antitrust Challenges to Category Captain Arrange-ments,” Journal of Public Policy & Marketing, 22 (Fall), 201–15.

Dhar, Sanjay, Stephen Hoch, and Nanda Kumar (2001), “EffectiveCategory Management Depends on the Role of the Category,”Journal of Retailing, 77 (2), 165–84.

Dreze, Xavier, Stephen Hoch, and Mary Purk (1994), “Shelf Man-agement and Space Elasticity,” Journal of Retailing, 70 (4), 301–26.

Du, Rex, Eunkyu Lee, and Richard Staelin (2005), “Bridge, Focus,Attack, or Stimulate: Retail Category Management Strategieswith a Store Brand,” Quantitative Marketing and Economics,3 (4), 393–418.

Fader, Peter, and Bruce G.S. Hardie (1996), “Modeling ConsumerChoice Among SKUs,” Journal of Marketing Research, 33(November), 442–52.

Federal Trade Commission (2001), “Report on the Federal TradeCommission Workshop on Slotting Allowances and OtherMarketing Practices in the Grocery Industry,” (February), https://www.ftc.gov/reports/report-federal-trade-commission-workshop-slotting-allowances-other-marketing-practices.

Federal Trade Commission (2003), “Antitrust and Category Cap-tains Roundtable Discussion,” (accessedMarch 10, 2016), http://www.antitrustinstitute.org/files/270.pdf.

Gelman, Andrew, and Jennifer Hill (2006), Data Analysis UsingRegression and Multilevel/Hierarchical Models. Cambridge,UK: Cambridge University Press.

Gooner, Richard, Neil Morgan, andWilliam Perreault Jr. (2011), “IsRetail Category Management Worth the Effort (and Does aCategory Captain Help or Hinder)?” Journal of Marketing,75 (September), 18–33.

Hanssens, Dominique, Leonard Parsons, and Randall Schultz(2003), Market Response Models: Econometric and Time SeriesAnalysis, Vol. 12. New York: Springer Science & Business Media.

Hoch, Stephen, Eric Bradlow, and Brian Wansink (1999), “TheVariety of an Assortment,” Marketing Science, 18 (4), 527–46.

Kadiyali, Vrinda, Pradeep K. Chintagunta, and Naufel Vilcassim(2000), “Manufacturer-Retailer Channel Interactions and Implica-tions for Channel Power: An Empirical Investigation of Pricingin a Local Market,” Marketing Science, 19 (2), 127–48.

Kok, Gurhan, and Marshall Fisher (2007), “Demand Estimation andAssortment Optimization Under Substitution: Methodology andApplication,” Operations Research, 55 (6), 1001–21.

Kumar, Nimalya, and Jan-Benedict Steenkamp (2007), PrivateLabel Strategy: How to Meet the Store Brand Challenge. Boston:Harvard Business Press.

Kurtulus, Mumin, and Alper Nakkas (2011), “Retail AssortmentPlanning Under Category Captainship,” Manufacturing & ServiceOperations Management, 13 (1), 124–42.

Kurtulus, Mumin, Alper Nakkas, and Sezer Ulku (2014), “TheValue of Category Captainship in the Presence of ManufacturerCompetition,” Production and Operations Management, 23 (3),420–30.

Kurtulus, Mumin, and Beril Toktay (2011), “Category Captainshipvs. Retailer Category Management Under Limited Retail ShelfSpace,” Production and Operations Management, 20 (1),47–56.

Kurtulus, Mumin, Sezer Ulku, Jeffrey P . Dotson, and Alper Nakkas(2014), “The Impact of Category Captainship on the Breadth andAppeal of a Retailer’s Assortment,” Journal of Retailing, 90 (3),379–92.

Lilien, Gary, Philip Kotler, and Sridhar Moorthy (1992), MarketingModels. New York: Prentice Hall.

Morgan, Neil, Anna Kaleka, and Richard Gooner (2007), “FocalSupplier Opportunism in Supermarket Retailer Category Man-agement,” Journal of Operations Management, 25 (2), 512–27.

Nijs, Vincent, Kanishka Misra, and Karsten Hansen (2014),“Outsourcing Retail Pricing to a Category Captain: The Role ofInformation Firewalls,” Marketing Science, 33 (1), 66–81.

Nowlis, Stephen, Ravi Dhar, and Itamar Simonson (2010), “TheEffect of Decision Order on Purchase Quantity Decisions,”Journal of Marketing Research, 47 (August), 725–37.

Pauwels, Koen, and Shuba Srinivasan (2004), “Who Benefits fromStore Brand Entry?” Marketing Science, 23 (3), 364–90.

Progressive Grocer (2010), “Smarter, Faster, Stronger,” 89 (10),42–92.

Progressive Grocer (2011), “Collective Excellence,” 90 (11),37–105.

Progressive Grocer (2015), “Winning Game Plans,” 94 (11),36–103.

Progressive Grocer (2016), “Selling Solutions,” 95 (11), 41–80.Rooderkerk, Robert, Harald Van Heerde, and Tammo Bijmolt

(2011), “Incorporating Context Effects into a Choice Model,”Journal of Marketing Research, 48 (August), 767–80.

Rooderkerk, Robert, Harald Van Heerde, and Tammo Bijmolt(2013), “Optimizing Retail Assortments,” Marketing Science,32 (5), 699–715.

Rossi, Peter, Greg Allenby, and Rob McCulloch (2005), BayesianStatistics and Marketing. Chichester, UK: John Wiley & Sons.

Subramanian, Upender, Jagmohan Raju, Sanjay Dhar, and YusongWang (2010), “Competitive Consequences of Using a CategoryCaptain,” Management Science, 56 (10), 1739–65.

Van Heerde, Harald, Peter Leeflang, and Dick Wittink (2004),“Decomposing the Sales Promotion Bump with Store Data,”Marketing Science, 23 (3), 317–34.

Wright, Joshua (2009), “Antitrust Analysis of Category Manage-ment: Conwood v United States Tobacco Co.,” Supreme CourtEconomic Review, 17 (1), 311–37.

Zenor, Michael (1994), “The Profit Benefits of Category Man-agement,” Journal of Marketing Research, 31 (May), 202–13.


https://www.ftc.gov/reports/report-federal-trade-commission-workshop-slotting-allowances-other-marketing-practices



http://www.antitrustinstitute.org/files/270.pdf

http://www.antitrustinstitute.org/files/270.pdf

yasin alan, jeffrey p. dotson, & m¨umin kurtulus¸ on the ... · category by developing new...

Documents